AI Glossary

An AI system that can autonomously perform tasks, make decisions, and take actions to achieve specified goals. Can use tools, access APIs, and operate with minimal human intervention.

Application Programming Interface that allows developers to access AI models and services. OpenAI, Anthropic, and others provide APIs to integrate their models into applications.

The simulation of human intelligence by machines, enabling them to perform tasks that typically require human cognition such as learning, reasoning, problem-solving, and decision-making.

A technique that allows models to focus on different parts of the input when producing outputs. Enables models to capture long-range dependencies in data.

Processing multiple AI requests together rather than one at a time. Can improve efficiency and reduce costs for non-time-sensitive workloads.

AI field focused on enabling machines to interpret and understand visual information from the world, including images and videos.

The maximum amount of text (measured in tokens) that a model can process at once. Larger context windows allow for more information to be considered when generating responses.

A subset of machine learning using neural networks with many layers (hence 'deep'). Particularly effective for complex tasks like image recognition and natural language processing.

Running AI models locally on devices (phones, IoT devices) rather than in the cloud. Offers benefits in latency, privacy, and offline operation.

A numerical representation of data (text, images, etc.) as vectors in a high-dimensional space. Similar items have similar embeddings, enabling semantic search and comparison.

Providing a model with a few examples of a task within the prompt to help it understand the desired output format and behavior.

The process of taking a pre-trained model and further training it on a specific dataset to customize it for particular tasks or domains.

AI systems that can create new content — text, images, code, music, video. Includes LLMs, image generators like DALL-E and Midjourney, and more.

When an AI model generates information that is factually incorrect or made up, but presents it as truth. A key challenge in deploying LLMs for factual applications.

The process of using a trained AI model to make predictions or generate outputs on new data. Distinct from training, which is the process of creating the model.

AI models trained on massive amounts of text data that can understand and generate human-like text. Examples include GPT-4, Claude, and Llama.

The time delay between sending a request to an AI model and receiving a response. Critical factor in user experience for real-time applications.

A subset of AI where systems learn and improve from experience without being explicitly programmed. ML algorithms use data to identify patterns and make predictions.

When an AI model's performance degrades over time as real-world data changes from what the model was trained on. Requires monitoring and periodic retraining.

AI systems that can process and generate multiple types of data — text, images, audio, video. GPT-4V and Gemini are examples of multimodal models.

The branch of AI focused on enabling computers to understand, interpret, and generate human language. Powers chatbots, translation, and text analysis.

A computing system inspired by biological neural networks in the brain. Consists of interconnected nodes (neurons) that process information in layers.

The practice of crafting effective inputs (prompts) for AI models to achieve desired outputs. Critical skill for working with LLMs.

A technique that combines LLMs with external knowledge retrieval. The model first retrieves relevant information from a knowledge base, then uses it to generate more accurate responses.

A training technique where models are refined based on human preferences and feedback. Used to make LLMs more helpful, harmless, and honest.

A parameter that controls randomness in model outputs. Higher temperature = more creative/random responses. Lower temperature = more focused/deterministic responses.

The basic unit of text that language models process. Can be a word, part of a word, or punctuation. Model pricing and limits are often measured in tokens.

A neural network architecture that revolutionized NLP. Uses attention mechanisms to process sequential data. Foundation of modern LLMs like GPT and BERT.

A database optimized for storing and querying embeddings (vectors). Essential for semantic search, recommendations, and RAG systems. Examples: Pinecone, Weaviate.

The ability of a model to perform tasks it wasn't explicitly trained for, using only natural language instructions. A key capability of modern LLMs.